Methylation markers for predicting sensitivity to treatment with antibody based therapy

Abstract

The present invention relates to a method of determining or predicting the sensitivity of a subject to an anti-inflammatory treatment against IBD using vedolizumab, comprising the steps of: Providing a biological sample of a subject suffering from IBD, determining the methylation status of at least one CpG selected from the group consisting of cg08081727, cg17830959, cg03455316, cg05197062, cg00441209, cg00706914, cg12906381, cg25299227, cg05338672, cg17764313, cg16467921, cg04674762, cg02601475, cg14115807, cg21070860, cg04546413, cg12667521, cg05062694, cg02229781, cg17096289, cg08017465, cg18319102, cg09659072, cg03161606, cg25267487, and determining the sensitivity based on said methylation status wherein a higher level of methylation of cg17830959, cg03455316, cg25299227, cg05197062, cg12906381, cg05338672, cg02601475, cg00706914, cg04674762, cg02229781, cg09659072, cg08017465, cg18319102, cg21070860, cg14115807, and a lower level of methylation of cg08081727, cg00441209, cg17764313, cg16467921, cg05062694, cg25267487, cg03161606, cg04546413, cg17096289, cg12667521 in comparison to a control value or control sample is indicative of an increased sensitivity to a therapy using vedolizumab.

Description

[0001] METHYLATION MARKERS FOR PREDICTING SENSITIVITY TO TREATMENT WITH ANTIBODY BASED THERAPY

[0002] TECHNICAL FIELD OF THE INVENTION

[0003] The present invention relates in general to the use of the methylation status of particular loci positioned on genomic DNA and their potential to predict drug resistance / susceptibility to the treatment of Crohn’s disease (CD) and ulcerative colitis (UC), referred to as inflammatory bowel diseases (IBD). More specifically, the invention relates to the DNA methylation status as biomarkers for predicting the efficacy of treatment using vedolizumab and ustekinumab in the treatment of CD.

[0004] BACKGROUND OF THE INVENTION

[0005] CD is an IBD with an unknown etiology. Based on advancing knowledge of the pathophysiology new therapeutic strategies have been developed that target the pathologic pathways in CD. Vedolizumab and (VDZ) and ustekinumab (USTE) other cytokines (IL-12 / IL-23) have demonstrated efficacy for treating CD. Biomarker discovery for CD based on the personalized genotype, clinical information and immunological reactivity could facilitate the stratification of patients to optimally tailored choice for therapeutic approaches

[0006] Despite the proven efficacy of vedolizumab (VDZ) or ustekinumab (USTE), many patients fail to respond or lose response overtime.. Therefore, predictive biomarkers for treatment success would be of extreme value. Previous studies associated differential DNA methylation with CD-specific phenotypes, suggesting a potential use in classification and prediction of response to treatment.

[0007] BRIEF DESCRIPTION OF THE FIGURES

[0008] Figure 1. Predictive model using stability selected gradient boosting for response to therapy. The orange and green boxes depict the machine learning set-up and model generation to predict response to vedolizumab and ustekinumab while the blue box shows the evaluation on multi -drug failure patients. A) Receiver operating characteristics plots showing the mean area under the curve (AUC) performance of the discovery and external validation cohorts. B) Radar plots presenting the difference in methylation between response (purple) and non-response (green) for the top 15 predictor CpG loci. C) Variable feature importance of the top 15 predictor CpG loci.

[0009] Figure 2. Longitudinal stability analyses. A) Volcanoplot representing the differential methylation analyses comparing into treatment with pretreatment where grey dots represent CpG loci located on the Illumina EPIC array and black dots represent response-associated predictor CpGs. X-axis represents mean difference in percentage methylation, Y -axis represents the statistical significance as depicted in -logio(p-value). B) Scatterplot showing the correlation of differential DNA methylation between R and NR pretreatment (Tl) and into treatment (T2). Grey dots represent CpG loci located on the Illumina EPIC array and black dots represent response-associated predictor CpGs. C) Intra-class correlation (ICC) coefficients of the predictor CpGs calculated when comparing pretreatment and into treatment as well as the ICC coefficients obtained from previous long-term stability analyses46. The vertical dashed grey lines represent classification boundaries introduced by Koo and Li103, with blocks representing poor (ICC < 0.5), moderate (0.5 < ICC < 0.75), good (0.75 < ICC < 0.9), and excellent (0.9 > ICC. D) Receiver operating characteristics plots representing the predictive performance into treatment (T2; black) and pre-treatment (Tl; grey) as reference.

[0010] Figure 3. Volcanoplot representing the differential methylation analyses comparing responders (R) with non-responders (NR) pretreatment for A) vedolizumab and B) ustekinumab. Grey dots represent CpG loci located on the Illumina EPIC array and black dots represent response-associated predictor CpGs. X-axis represents mean difference in percentage methylation, Y-axis represents the statistical significance as depicted in -logio(p-value). Dumbbell plot representing the -loglO(p-value) before (blue) and after (red) correcting for age, sex, and estimated cellular composition for C) vedolizumab and D) ustekinumab. Transparent dots represent differences with a p-value above 0.05, while full dots represent differences with a p-value < 0.05.

[0011] Figure 4 A) receiver operating characteristics plots showing the mean AUC performance of the VDZ and USTE models in the subsets of validation cohort patients that were assessed using the strict (blue) and modified (pink) definitions of response. B) receiver operating characteristics plots showing the mean AUC performance of the VDZ and USTE models in the subsets of validation cohort patients that were anti-TNF exposed (black) or non-exposed (yellow).

[0012] SUMMARY OF THE INVENTION

[0013] The invention provides a method of determining or predicting the sensitivity of a subject to an antiinflammatory treatment against IBD using vedolizumab, comprising the steps of:

[0014] Providing a biological sample of a subject suffering from IBD, determining the methylation status of at least one CpG selected from the group consisting of cgl7830959, cg21070860, cg08081727, cg02601475, cg00706914, cgl7096289, cg25299227, cg05197062, cg00441209, cg03455316, cg05338672, cgl266752I, cg05062694, cgl7764313, cg09659072, cg04674762, cg02229781, cg25267487, cgl6467921, cg08017465, cgl4115807, cgl8319102, cgl2906381, cg04546413, cg03161606, and determining the sensitivity based on said methylation status wherein a higher level of methylation of cgl7830959, cg21070860, cg02601475, cg00706914, cg25299227, cg05197062, cg03455316, cg05338672, cg09659072, cg04674762, cg02229781, cg08017465, cgl4115807, cgl8319102, cgl2906381, and a lower level of methylation of cg08081727, cgl7096289, cg00441209, cgl266752I, cg05062694, cgl7764313, cg25267487, cgl6467921, cg04546413, cg03161606 in comparison to a control value or control sample is indicative of an increased sensitivity to a therapy using vedolizumab.

[0015] The invention further provides a method of determining or predicting the sensitivity of a subject to an anti-inflammatory treatment against IBD using Ustekinumab, comprising the steps of: Providing a biological sample of a subject suffering from IBD, determining the methylation status of at least one CpG selected from the group consisting of cgl3982436, cgl 1079896, cg09147516, cgl9162470, cg05541470, cgl3754978, cgl8771300, cgl3816228, cg08270491, cgl4578009, cg01966334, cg02265440, cgl7190362, cgl3468767, cgl7132030, cg204345I I, cg23973310, cg25279747, cgl l747594, cgl7481116, cg02086964, cgl 1141652, cgl2119625, cgl5921713, cg21650737, cgl4829155, cg07620573, cg25535666, cg02860608, cg20707527, cgl4157578, cgl7037048, cg27408471, cg04996388, cg22635676, cg24309011, cg08359343, cg04043455, cgl6677969, cg06546677, cgl7851604, cgl4024893, cgl9072128, cg22329875, cgl 1935738, cg08010094, cg23264413, cgl 1146691, cgl3746813, cg05303293, cg24536782, cgl0167378, cg23762517, cg08109568, cgl l787544, cg00876837, cg06880335, cgl9678447, cg08017465, cg08993878, cg02094681, cg20171775, cgl0403394, cg08777654, cgl3679714, cg09125754, cg23730027, cgl0864200, and determining the sensitivity based on said methylation status wherein a higher level of methylation of cgl 1079896, cg05541470, cgl3754978, cgl8771300, cgl4578009, cg02265440, cgl7190362, cgl3468767, cg25279747, cgl 1747594, cgl7481116, cg02086964, cgl5921713, cg21650737, cgl4829155, cg07620573, cg02860608, cgl4157578, cgl7037048, cg27408471, cg22635676, cg24309011, cg08359343, cg04043455, cgl7851604, cgl4024893, cgl9072128, cg22329875, eg 11146691, cg24536782, cg08109568, egl 1787544, cgl9678447, cg20171775, cg08777654, cgl3679714, and a lower level of methylation of cgl3982436, cg09147516, cgl9162470, cgl3816228, cg08270491, cg01966334, cgl7132030, cg20434511, cg23973310, egl 1141652, cgl2119625, cg25535666, cg20707527, cg04996388, cgl6677969, cg06546677, egl 1935738, cg08010094, cg23264413, cgl3746813, cg05303293, cgl0167378, cg23762517, cg00876837, cg06880335, cg08017465, cg08993878, cg02094681, cgl0403394, cg09125754, cg23730027, eg 10864200 in comparison to a control value or control sample is indicative of an increased sensitivity to a therapy using Ustekinumab.

[0016] Preferably, the method are as described above, wherein said IBD is Crohn’s disease. Preferably, wherein said biological sample comprises white blood cells. Preferably, said methylation level is determined using DNA methylation array.

[0017] Preferably, said method of determining or predicting the sensitivity of a subject to an antiinflammatory treatment against IBD using vedolizumab according to the invention, wherein at least one CpG comprises the first 2, more preferably the first 3, 4, 5, 6, 7. 8. 9. 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or all CpGs of Table 3. In an embodiment, said CpG comprises the CpGs listed in Figure IB. In an embodiment, said CpG comprises the CpGs listed in Figure 1C. In an embodiment, said CpG comprises the CpGs listed in Figure 3C. In another embodiment, said at least one CpG comprises at least 2 3, 4, 5, 6, 7. 8. 9. 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or all CpGs of Table 3. The more CpGs included, the better the predictive value of the marker set.

[0018] Preferably, said method of determining or predicting the sensitivity of a subject to an antiinflammatory treatment against IBD using Ustekinumab according to the invention, wherein at least one CpG comprises the first CpG first 2, more preferably the first 3, 4, 5, 6, 7. 8. 9. 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67 or all CpG of Table 4. In an embodiment, said CpG comprises the CpGs listed in Figure 1C. In an embodiment, said CpG comprises the CpGs listed in Figure 3D. In another embodiment, said at least one CpG comprises at least 4, 5, 6, 7. 8. 9. 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67 or all CpG of Table 4. The more CpGs included, the better the predictive value of the marker set.

[0019] The invention further provides Vedolizumab for use in the treatment of IBD in a subject who is sensitive to a treatment with vedolizumab as determined using the method determining or predicting the sensitivity of a subject to an anti-inflammatory treatment against IBD using vedolizumab according to the invention. The invention further provides Ustekinumab for use in the treatment of IBD in a subject who is sensitive to a treatment with vedolizumab as determined using the method determining or predicting the sensitivity of a subject to an anti-inflammatory treatment against IBD using Ustekinumab according to the invention.

[0020] DETAILED DESCRIPTION OF THE INVENTION

[0021] Definitions

[0022] The term “differentially methylated region(s)” (DMRs) as used herein refer to genomic regions with different DNA methylation status across different biological samples and regarded as possible functional regions involved in gene transcriptional regulation.

[0023] The term “biological sample” as used herein refers to any sample from an IBD patient for diagnostic, prognostic, or personalized medicinal uses and may be obtained from surgical samples, such as biopsies or fine needle aspirates, from paraffin-embedded tissues, from a fresh or frozen body fluid. Most preferably the sample contains nucleated blood cells. However, any other suitable biological samples (e.g. bodily fluids such as stool, etc...) in which the methylation status of a locus of interest can be determined are included within the scope of the invention.

[0024] By “methylation status” is meant the level of methylation of cytosine residues (found in CpG pairs) in the DNA sequence of interest. When used in reference to a CpG site, the methylation status may be methylated or unmethylated. When used in reference to a CpG island or to any stretch of residues, the methylation status refers to the level of methylation, which is the relative or absolute concentration of methylated C at the particular CpG island or stretch of residues in the DNA present in a biological sample.

[0025] The term "hypermethylation" as used herein refers to the average methylation status corresponding to an increased presence of 5-mCyt at one or a plurality of CpG dinucleotides within a DNA sequence of a test DNA sample, relative to the amount of 5-mCyt found at corresponding CpG dinucleotides from a DNA sample from a non-responder. Preferably, a responder is a subject who meets the following criteria at their follow-up endoscopy between 6-24 months after starting biological treatment (if no major medication changes during this period): a) - >50% drop in the endoscopic severity score SES-CD score compared to baseline endoscopy, and b) - >3 point drop in the clinical Harvey Bradshaw index (51) and / or c) - shows signs of biological improvement based on either a CRP <5.0 mg / 1 or fecal calprotectin <250ug / g, and d) - need no steroid treatment at response assessment. A non- responder is preferably a subject who meets the following criteria at their follow-up endoscopy between 6-24 months after starting biological treatment (if no major medication changes during this period): a)- <50% drop in SES-CD score compared to baseline endoscopy, and b) - <3 point drop in Harvey Bradshaw index, and / or c) - shows no signs of biological improvement based on either an unchanged or increased (delta 25%) CRP and / or fecal calprotectin >250ug / g, and / or d) - is in need of a steroid treatment at response assessment.

[0026] The phrase "corresponding to" when used to describe positions or sites within nucleotide sequences, is used herein as it is understood in the art. As is well known in the art, two or more nucleotide sequences can be aligned using standard bioinformatic tools, including programs such as BLAST, ClustalX, Sequencher, and etc. Even though the two or more sequences may not match exactly and / or do not have the same length, an alignment of the sequences can still be performed and, if desirable, a "consensus" sequence generated. Indeed, programs and algorithms used for alignments typically tolerate definable levels of differences, including insertions, deletions, inversions, polymorphisms, point mutations, etc. Such alignments can aid in the determination of which positions in one nucleotide sequence correspond to which positions in other nucleotide sequences.

[0027] The abbreviation "CpG" is used herein to refer to a dinucleotide comprised of a cytosine nucleotide (deoxycytidine) linked via a phosphate group to a guanine nucleotide (deoxyguanosine) through linkages to the 5' position of the deoxy cytidine and the 3' position of the deoxyguanosine. The cytosine in this dinucleotide is said to be in the "5' position" of the dinucleotide, and the guanine is said to be in the "3' position" of the dinucleotide. As is understood by one of ordinary skill in the art, the abbreviation "CpG" also refers to modified dinucleotides similar, so long as the 5' nucleotide is still identifiable as deoxycytidine and the 3' nucleotide is still identifiable as deoxyguanosine. For example, a deoxycytidine-deoxyguanosine dinucleotide in which the cytosine ring is methylated at the 5 position is still considered a CpG dinucleotide, and may be referred to as a methylated CpG or abbreviated as 5mCpG. As used herein, the abbreviation CpG can also refer to a CpG site, defined below.

[0028] The term “CpG island” refers to a contiguous region of genomic DNA that satisfies the criteria of (1) having a frequency of CpG dinucleotides corresponding to an “Observed / Expected Ratio” >0.6, and (2) having a “GC Content” >0.5. CpG islands are typically, but not always, between about 0.2 to about 1 kb in length.

[0029] The term “Observed / Expected Ratio” (“O / E Ratio”) refers to the frequency of CpG dinucleotides within a particular DNA sequence, and corresponds to the (number of CpG sites / (number of C bases X number of G bases)) X band length for each fragment.

[0030] The term "CpG site" is used herein to refer to a position within a region of DNA corresponding to a position where a CpG dinucleotide is found in a reference sequence. One of ordinary skill in the art will understand the term CpG site to encompass the location in the region of DNA where a CpG dinucleotide is typically found, whether or not the dinucleotide at that position is a CpG dinucleotide in a particular DNA molecule. For examples, the DNA sequence of a gene may typically contain a CpG dinucleotide at a particular position, but may contain other dinucleotides at the corresponding position in mutant versions, polymorphic variants, or other variations of the gene. Some mutations such as single base substitutions may alter the identity of the dinucleotide to be something other than CpG. Other mutations such as insertions or deletions may alter the position of the CpG dinucleotide typically found at a particular site. In cases such as these, the term CpG site is understood by one of ordinary skill in the art to encompass the site corresponding to the position where a CpG dinucleotide is typically found, for example, in individuals not carrying a mutation at this location. Similarly, the term CpG site also encompasses the corresponding site in a nucleic acid that has been modified experimentally, for example by labeling, methylation, demethylation, deamination (including conversion of a cytosine to uracil by a chemical such as sodium bisulfite), etc.

[0031] The term "predicting the sensitivity to a treatment", as used herein refers to the determination of the likelihood that the patient will respond either favorably or unfavorably to a given therapy. Especially, the term "prediction", as used herein, relates to an individual assessment of any parameter that can be useful in determining the evolution of a patient. Preferred confidence intervals are at least 50%, at least 60%, at least 70%, at least 80%, at least 90% at least 95%. The p- values are, preferably, 0.2, 0.1 or 0.05.

[0032] Detailed description of certain embodiments of the invention

[0033] Methylation assays

[0034] The methylation status of a DMR in the DNA of a biological sample may be determined using any methylation assay. Several quantitative methylation assays are commercially available. These include COBRA™ (Ziong and Laird, Nucleic Acid Res 1997 25; 2532-4) which uses methylation sensitive restriction endonuclease, gel electrophoresis and detection based on labeled hybridization probes. Another available technique is the Methylation Specific PCR (MSP) for amplification of DNA segments of interest. This is performed after sodium 'bisulfite' conversion of cytosine using methylation sensitive probes. Methy Light™, a quantitative methylation assay-based uses fluorescence based PCR (Eads et al, Cancer Res 1999; 59:2302-2306). Another method used is the Quantitative Methylation (QMTM) assay, which combines PCR amplification with fluorescent probes designed to bind to putative methylation sites. Ms-SNuPE™ is a quantitative technique for determining differences in methylation levels in CpG sites. As with other techniques bisulfite treatment is first performed leading to the conversion of unmethylated cytosine to uracil while methyl cytosine is unaffected. PCR primers specific for bisulfite converted DNA is used to amplify the target sequence of interest. The amplified PCR product is isolated and used to quantitate the methylation status of the CpG site of interest (Gonzalgo and Jones Nuclei Acids Res 1997; 25:252-31). The preferred method of measurement of cytosine methylation is the Illumina method.

[0035] Illumina method

[0036] For DNA methylation assay the Illumina Infmium® Infmium MethylationEPIC Beadchip assay is preferably used for genome wide quantitative methylation profding. Briefly genomic DNA is extracted from cells in this case whole blood, for which the original source of the DNA is preferably white blood cells. Using techniques widely known in the trade, the genomic DNA may be isolated using commercial kits. Proteins and other contaminants are preferably removed from the DNA, preferably using proteinase K. The DNA may be removed from the solution using available methods such as organic extraction, salting out or binding the DNA to a solid phase support.

[0037] Bisulfite Conversion

[0038] As described in the Infmium® Assay Methylation Protocol Guide, prior to its analysis, DNA must be treated with sodium bisulfite which converts unmethylated cytosine to uracil, while the methylated cytosine remains unchanged. The bisulfite converted DNA is then denatured and neutralized. The genomic DNA may be bisulfite converted using commercial kits, e.g ZYMO®.

[0039] The denatured DNA is then amplified. The whole genome application process increases the amount of DNA by up to several thousand-fold. The next step uses enzymatic means to fragment the DNA. The fragmented DNA is next precipitated using isopropanol and separated by centrifugation. The separated DNA is next suspended in a hybridization buffer. The fragmented DNA is then hybridized to beads that have been covalently limited to 50 mer nucleotide segments at a locus specific to the cytosine nucleotide of interest in the genome. There are a total of over 850,000 bead types specifically designed to anneal to the locus where the particular cytosine is located. The beads are bound to silicon based arrays. There are two bead types designed for each locus, one bead type represents a probe that is designed to match to the methylated locus at which the cytosine nucleotide will remain unchanged. The other bead type corresponds to an initially unmethylated cytosine which after bisulfite treatment is converted to a thiamine nucleotide. Unhybridized (not annealed to the beads) DNA is washed away leaving only DNA segments bound to the appropriate bead and containing the cytosine of interest. The bead bound oligomer, after annealing to the corresponding patient DNA sequence, then undergoes single base extension with fluorescently labeled nucleotide using the 'overhang' beyond the cytosine of interest in the patient DNA sequence as the template for extension.

[0040] If the cytosine of interest is unmethylated then it will match perfectly with the unmethylated or "U" bead probe. This enables single base extensions with fluorescent labeled nucleotide probes and generate fluorescent signals for that bead probe that can be read in an automated fashion. If the cytosine is methylated, single base mismatch will occur with the "U" bead probe oligomer. No further nucleotide extension on the bead oligomer occurs however thus preventing incorporation of the fluorescent tagged nucleotides on the bead. This will lead to low fluorescent signal form the bead "U" bead. The reverse will happen on the "M" or methylated bead probe.

[0041] Laser is used to stimulate the fluorophore bound to the single-base used for the sequence extension. The level of methylation at each cytosine locus is determined by the intensity of the fluorescence from the methylated compared to the unmethylated bead. Cytosine methylation level is expressed as "W which is the ratio of the methylated-bead probe signal to total signal intensity at that cytosine locus. These techniques for determine cytosine methylation have been previously described and are widely available for commercial use.

[0042] In preferred embodiments of the invention, the detected methylation status is hypermethylation, an increase in the methylation of cytosines in CpG sites in responders compared to non-responders. The increase can be about 1%, about 5%, about 10%, about 20%, about 25%, about 30%, about 35%, about 40%, about 45%, about 50%, about 60%, about 65%, about 70%, about 75%, about 80%, about 85%, about 90%, about 95%, or more than about 95% greater than the extent of methylation of cytosines typically expected or observed for the CpG site or sites being evaluated.

[0043] In preferred embodiments of the invention, the detected methylation status is hypomethylation, a decrease in the methylation of cytosines in CpG sites in responders compared to non-responders. The decrease can be about 1%, about 5%, about 10%, about 20%, about 25%, about 30%, about 35%, about 40%, about 45%, about 50%, about 60%, about 65%, about 70%, about 75%, about 80%, about 85%, about 90%, about 95%, or more than about 95% less than the extent of methylation of cytosines typically expected or observed for the CpG site or sites being evaluated.

[0044] Suitable controls may need to be incorporated in order to ensure the method chosen is working correctly and reliably. Suitable controls may include assessing the methylation status of a gene known to be methylated. This experiment acts as a positive control to ensure that false negative results are not obtained. The DMR may be one which is known to be methylated in the sample under investigation or it may have been artificially methylated. In one embodiment, the DMR of interest may be assessed in normal cells, following treatment with Sssl methyltransferase, as a positive control. Additionally or alternatively, suitable negative controls may be employed with the methods of the invention. Here, suitable controls may include assessing the methylation status of a gene known to be unmethylated or a DMR that has been artificially demethylated. This experiment acts as a negative control to ensure that false positive results are not obtained. In one embodiment, the DMR of interest may be assessed in normal cells as a negative control, in particular if the DMR is unmethylated in normal tissues.

[0045] In a preferred embodiment, the method according to the invention is determined, by determining if hyper- or hypomethylation is determined in at least two, preferably three, four or five DMRs as disclosed herein. An advantage thereof is that if more of said DMRs are hyper or hypomethylated, the sensitivity of the method is greater.

[0046] In certain embodiments of the invention, the detected methylation status is determined in part or wholly on the basis of a comparison with a control. The control can be a value or set of values related to the extent and / or pattern of methylation in a control sample. In certain embodiments of the invention, such a value or values may be determined, for example, by calculations, using algorithms, and / or from previously acquired and / or archived data. In certain embodiments of the invention, the value or set of values forthe control is derived from experiments performed on samples or using a subject. For example, control data can be derived from experiments on samples derived from biological samples form subjects known to be sensitive to an anti-inflammatory treatment against IBD, using a monoclonal antibody targeting tumor necrosis factor a or integrin a4p7.

[0047] The term ’’control value” as used herein refers to a value representing the quantity of methylation of said DMR in a non-responder (e.g., from white blood cells from a non-responder), thereby quantifying the average methylation density in the locus compared to the methylation density of the control DNA. A skilled person will understand that it is also possible to compare the methylation of a certain DMR with the methylation level of the same DMR in a responder. In such case, when the methylation status is at a similar level compared to a control representative of a responder (which could be increased or decreased compared to a non-responder), this indicates that a subject is sensitive to a treatment. Using control values of methylation in DMRs of responders is also within the scope of this invention.

[0048] In certain embodiments of the invention, a control comprising DNA that is mostly or entirely demethylated, at one or more of the CpG sites being analyzed, is used. Such a control might be obtained, for example, from mutant tissues or cells lacking methyltransferase activity and / or from tissues or cells that have been chemically demethylated. For example, controls may be obtained from tissues or cells lacking activity of methyltransferases Dnmtl, Dnmt2, Dnmt3a, Dnmt3b, or combinations thereof. Agents such as 5 -aza-2'-deoxy cytidine may be used to chemically demethylate DNA.

[0049] In certain embodiments of the invention, a control comprising DNA that is mostly or entirely methylated, at one or more of the CpG sites being analyzed, is used. Such a control might be obtained, for example, from cells or tissues that are known or expected to be mostly or entirely methylated at the CpG site or sites of interest. Such a control could also be obtained by cells or tissues in which methylation levels have been altered and / or manipulated, for example, by overexpression of methyltransferases (such as enzymes Dnmtl, Dnmt2, Dnmt3a, Dnmt3b, any of the bacterial 5 -CpG methyltransferases, or combinations thereof). In certain embodiments of the invention, samples used to obtain control values are processed and / or manipulated in the same manner as the samples being evaluated. In a preferred embodiment, said control comprises a colorectal cancer cell line with a known methylation status.

[0050] In some embodiments, negative control samples are generated by treating positive control samples with a demethylating agent, preferably 5 -aza-2'-deoxy cytidine.

[0051] Medical treatment using VDZ In another aspect, the invention provides VDZ for use in the treatment of a subject who is sensitive to a treatment with VDZ in the treatment as determined using the method according to the invention. If a positive clinical response to treatment with VDZ is determined, the patient is identified or selected for a treatment with VDZ. If a negative clinical response to a treatment is determined, the patient is not selected for treatment, and one or more alternative drug or medical intervention may be more beneficial for the treatment of IBD, preferably CD.

[0052] Medical treatment using USTE

[0053] In another aspect, the invention provides USTE for use in the treatment of a subject who is sensitive to a treatment with USTE in the treatment as determined using the method according to the invention. If a positive clinical response to treatment with USTE is determined, the patient is identified or selected for a treatment with USTE. If a negative clinical response to a treatment is determined, the patient is not selected for treatment, and one or more alternative drug or medical intervention may be more beneficial for the treatment of IBD, preferably CD.

[0054] The above disclosure generally describes the present invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as those commonly understood by one of ordinary skill in the art to which this invention belongs. A more complete understanding can be obtained by reference to the following specific examples which are provided herein for purposes of illustration only, and are not intended to limit the scope of the invention.

[0055] EXAMPLE

[0056] Study population description

[0057] To identify epigenetic signatures that associate with objective response to VDZ and USTE, we prospectively recruited a discovery cohort of 126 adult CD patients with active clinical- (median Harvey Bradshaw Index (HBI) 8 (IQR 4-12)), biochemical- (median C-reactive protein (CRP) 6.1 mg / L (IQR 2.2-14.8); median fecal calprotectin (FCP) 903 pg / g (IQR 278-1816)) and endoscopic (median simple endoscopic score-CD (SES-CD) 9 (IQR 6-15)) disease activity at the Amsterdam University Medical Centers. In addition, an external validation cohort of 58 adult CD patients were recruited at the John Radcliffe Hospital, Oxford, United Kingdom. Peripheral blood leukocyte (PBL) samples were obtained prior to [Tl] and during response assessment [T2], where patients were defined as responder (R) or non-responder (NR). An overview of all the clinical characteristics across the different cohorts and treatments can be found in Table 1.

[0058] The VDZ discovery cohort consisted of a total of 64 patients (NR=36, NNR=28) of which 49 (77%) were previously exposed to anti-TNF treatment and nine (14%) to USTE. R and NR presented overall comparable clinical characteristics, without significant differences in age, sex and smoking behavior (p=0.98, 0.17, and 0.46 respectively). NR were more often exposed to anti-TNF (89.3% vs. 66.7%, p=0.03) and USTE (25% vs. 5.6%, p=0.03) compared to R. In addition, while a higher proportion of NR received an extra infusion (35.7% vs 13.9%, p=0.04) during follow-up at week 10, both the distribution of treatment interval (p=0.26) and serum VDZ concentrations at T2 were not significantly different between R and NR (median 15 (IQR 7.7-20.5) versus median 14 (IQR 3.6-27.5), p=0.77). Notably, 13 out of the 15 anti-TNF naive patients (87%) that started VDZ as first-line biological were classified as R and presented a shorter disease duration (p=0.02), lower rates of previous surgery (p=0.003) and perianal disease (p=0.004) as well as a higher percentage of Bl phenotype (p=0.01) compared to anti-TNF experienced patients.

[0059] The VDZ validation cohort consisted of 25 patients (NR=14 and NNR=11) with a median 9 (IQR 3-15) year disease duration of which 17 (68%) were assessed using a modified definition of response (see methods). Out of these 25 patients, 13 were biological naive (52%), 12 (48%) were anti-TNF- experienced and 7 (28%) were previously treated with USTE [enriched among the NR group (54.5% vs 7.1%, p=0.01)].). All other clinical characteristics were comparable in R and NR, including age, sex, and smoking behavior (p=0.56, 0.42 and 0.30 respectively). Similar to the discovery VDZ cohort, we noticed that among the biological naive patients the majority (69%) were R. The USTE discovery cohort consisted of 62 patients (NR=30, NNR=32) of which 61 (98%) were previously exposed to anti-TNF and 26 (42%) to VDZ. Overall clinical characteristics did not significantly differ between R and NR. While no significant differences in age and smoking behavior were observed between R and NR (p=0.80 and 0.90), the R population consisted of more female patients (R = 80% ; NR = 56.3%, p=0.05). No significant differences in treatment intensification (21.9% vs 6.7%, p=0.08), extra intravenous boost infusions (p=0.26) or serum USTE concentrations at T2 between R and NR were observed (median 3.0 (IQR 1.8-5.5) versus median 4.8 (IQR 2.1-8.6), p=0.31).

[0060] The USTE validation cohort consisted of 33 patients (NR=22 and NNR=11), with a median disease duration of 11 (IQR 3-20) years of which 25 (76%) were assessed using a modified definition of response. Of the 33 included patients, 9 were biological naive (52%), 21 (63.6%) were anti-TNF- experienced and 5 (15.2%) were previously treated with VDZ. R presented a significantly longer disease duration compared to NR (median 15 versus 5 years, p=0.05). No significant differences in age, sex, and smoking behavior (p=0.25, 0.63 and 0.24 respectively) were observed.

[0061] Blood DNA methylation profile predicts response to vedolizumab and ustekinumab

[0062] To identify prognostic biomarkers of VDZ and UST response we performed supervised machine learning through stability selected gradient boosting on blood samples obtained pre-treatment [Tl] to classify R from NR on the discovery cohort samples (Figure 1A)31. We were able to predict response with an area under the receiver operator curve (AUROC) of 0.87 and 0.89 (Figure IB) using an epigenetic signature of 25 and 68 CpGs for VDZ and USTE, respectively (Figure 1C-E). As validation, we tested our models against the external validation cohorts (Figure 1A). Indeed, our predictive models performed well with an AUROC of 0.75 for both VDZ and USTE (Figure IB), indicating reproducible response-associated differences in DNA methylation that precede treatment.

[0063] Focusing on the practical implications within a clinical setting, we computed the accuracy, likelihood ratio, and post-test likelihood for both models. These measures aid clinicians in the probability of an accurate response following a positive test outcome, considering the performance of the model, alongside the present-day probability of response to each drug. The recall and precision values of the VDZ-model result in an accuracy of 73%, with a likelihood ratio of 2.48 and a post-test likelihood of 67% given the currently reported week 52 endoscopic response of 45%5. In comparison, the USTE model achieves an accuracy of 83%, likelihood ratio of 4.58, and a post-test likelihood of 77% given the currently reported week 52 endoscopic response of 42% (Table 2)6. As the validation cohort consisted of samples whose response was defined on both strict and less strict criteria, we investigated whether a difference in performance could be observed. Stratifying the prediction scores indicated a notably higher performance when using a combination of clinical- and endoscopic endpoints for both VDZ (AUROCstnct=0.83 versus AUROCmodified=0.66) and USTE (AUROCstrict=0.83 versus AUROCmodified=0.72). Moreover, an increasingly common clinical scenario is the choice of treatment in patients who have already been exposed to anti-TNF therapies. Accordingly, we were interested whether the performance of our models was affected by the prior experience to anti-TNF medication. We observed a better performance among anti-TNF naive patients for both VDZ (AUROCeXposed=0.66 versus AUROCnOn-exposed=0.85) and USTE (AUROCeXposed=0.63 VerSUS AUROCnon-exposed=0.97).

[0064] Methylation of predictor CpGs does not change over the course of treatment

[0065] As samples were acquired pre-treatment [Tl], we sought to understand whether the start of treatment affected the methylation of the predictor CpGs. To this end, we compared samples obtained during response assessment [T2] with samples obtained pre-treatment [Tl], This approach indicated no statistically significant differences in DNA methylation for any of the predictor CpGs (Figure 2A). Indeed, comparing the differences over time suggested that the mean difference between R and NR was similar both pre- and into treatment (Figure 2B). Furthermore, a two-way, mixed, consistency intraclass correlation (ICC) analyses indicated stability to highly-stable DNA methylation over time with 24 out of 25 VDZ and 62 out of 68 USTE predictor CpGs presenting ICC values >0.75 (Figure 2B ). This observation was corroborated by interrogating our previous longitudinal consistency analysis32, where we observed that 16 out of 25 VDZ and 52 out of 68 USTE predictor CpGs presented ICC values >0.75 over a median span of 7 years (Figure 2C). As a final validation, we utilized the prognostic model to predict response to therapy of the samples obtained into treatment, where we obtained better performances compared to the pre-treatment samples (AUROCVDZ = 0.97; AUROCUSTE = 1.00) (Figure 2D). Altogether, our observations suggest that response- associated differences in DNA methylation present prior to treatment remain constant during treatment.

[0066] Combination of prediction models accurately identifies non-response in an independent cohort of patients that had failed multiple drug modalities

[0067] Having established that the prognostic response prediction models for both drugs perform well both pre- and into treatment, we explored whether our model could work on a retrospective cohort of 34 CD patients whose blood had been collected after failing a combination of anti-TNF, VDZ, and / or USTE. Treatment failure was determined through endoscopy in 28 (82%), MRI in 2 (6%), or the need for surgery in 4 (12%), combined with CRP in 35% and / or FCP in 16 (47%) and clinical outcome in 32 (94%) respectively. We observed a relatively large number of patients with extensive disease location (67.6%), perianal disease (52.9%) and had received IBD-related surgery (70.6%) in the past (Table 5).

[0068] Both the VDZ- and USTE response prediction models accurately predicted NR, with 24 out of 27 (88.9%) and 24 out of 26 (92.3%) patients being correctly flagged as NR to VDZ and USTE, respectively. Focusing specifically on the patients that failed both VDZ and USTE correctly identified 16 out of 19 (84%) as NR. Notably, 12 of out of these 19 patients (75%) were anti-TNF experienced.

[0069] Differentially methylated signal is not associated with common confounding factors

[0070] Through multiple linear regression analyses we confirmed the linearly separable nature for 22 of the 25 VDZ response-associated CpGs and 38 of the 68 USTE response-associated CpGs (Figure 3A and B). This difference in USTE response-associated CpGs indicates that a more complex non-linear relationship probably exists among the response-associated predictor CpGs. As it has been established that the peripheral blood DNA methylome is associated with common demographic factors such as sex, age, CRP, fecal calprotectin, and smoking status, as well as the underlying cellular composition33'36, we next investigated whether the aforementioned factors confounded our results. By correcting for these covariates in our linear regression analysis, we observed that 12 and 30 remained significantly associated with response for VDZ and USTE, respectively (Figure 3C and D). Our results thereby indicate that the response-associated predictor CpGs are likely independent of sex, age, smoking behavior and changes in the blood cell composition. We next also investigated whether our predictor CpG’s were significantly associated with CRP and fecal calprotectin, as has previously been reported by Somineni et al37. Accordingly, for VDZ 5 predictor CpGs significantly associated with CRP whereas a single CpG associated with fecal calprotectin. For USTE, we observed 9 predictor CpGs to associated with CRP and 2 predictor CpG with fecal calprotectin. Our observations indicate that the majority of the identified biomarker panels are independent of baseline inflammatory status. VDZ response is characterized by differential methylation of MHC Class I whereas USTE response appears more diffuse

[0071] We next investigated what genes were associated to the predictor CpGs. We observed that 20 of the 25 VDZ response-associated predictor CpGs were annotated to 16 unique genes (Table 3). Gene Ontology (GO) overrepresentation analyses identified enrichment for the MHC class I-related GO-terms (Figure 4A). Further interrogation of HLA-C indicated a regional difference in DNA methylation (Figure 4B). For USTE we found that 54 of the 68 UST predictor CpGs annotated to known 52 unique genes (Table 4). GO-term overrepresentation analyses did not identify a particular pathway that we could relate to immunological functions with most GO-terms being involved in general housekeeping (Figure 4C). In addition, we noticed that several predictor CpGs were identified within genes encoding RHOJ a gene encoding a GTP -binding protein (Figure 4D).

[0072] Discussion

[0073] Although the introduction of biologicals has sparked a revolution in the clinical care of CD, the current method of treatment selection remains suboptimal. Given the absence of randomized head- to-head trials, the current body of real-world data, which has yielded inconclusive results, indicates the need for an individualized approach38-39. Although ongoing drug development provides novel treatment options, predictive biomarkers that allow for pre-selection of successful treatment to the currently available biologicals, are urgently needed.

[0074] Here, we conducted a longitudinal case-control study where we identified 25 and 68 epigenetic- markers that prognostically predict objective response to VDZ and USTE, respectively, in a large cohort of strictly phenotyped CD patients. While both models were primarily developed using combined endoscopic- and clinical response indices, we demonstrate good performance of these biomarker panels against an independent, external validation in which 72.4% were assessed using a modified definition of response, suggesting overall generalizability of our results to a much larger IBD population. Notably, the subset of validation patients that were phenotyped according to our strict combined endpoint enhanced the performance of both the VDZ and USTE models, reinforcing the credibility of both models in identifying objective responders to biological therapy.

[0075] Furthermore, recent real-world data and post-hoc findings from the GEMINI trial indicate a superior response to VDZ in anti-TNF naive patients, however this outcome appeared less conclusive for USTE40-42. Although in our discovery cohorts, 77% of patients for VDZ and 98% for USTE were previously exposed to an anti-TNF, stratifying the validation patients by previous anti-TNF exposure remarkably showed that both models performed better in anti-TNF naive- rather than exposed patients. However the number of patients included in both subset comparisons are relatively small and further exploration using larger groups patients are needed.

[0076] Moreover, we demonstrate the ability of our models to effectively identify patients that previously failed both VDZ and USTE treatment. This holds true for both anti-TNF naive- and experienced patients, providing significant importance for clinical practice, as the precise prediction of non-response to both drugs offers clinicians the opportunity to make informed decisions. For anti- TNF experienced patients, this could mean a direct choice towards newer modes-of-action (i.e. JAK- inhibitors or SIP modulators). Conversely, for anti-TNF naive patients, it suggests a higher likelihood of responding positively to anti-TNF treatments, before moving on towards novel modes-of-action.

[0077] While this study is the first to demonstrate the theranostic utility of DNA methylation profiling for objective response to VDZ and USTE in CD patients, previous analyses from two separate studies explored its application to predict anti-TNF response in both CD and UC patients. In the first study, the authors sought to identify an anti-TNF response-associated profile combining integrated methylation and gene expression data from samples taken before and 2 weeks into treatment using a primary endpoint of clinical remission at week 1443. The observations were made using a cohort of 37 IBD (18 CD, 19 UC) patients and were validated using a publically available gene expression dataset containing 20 CD patients. In the second study, the authors do not show replication of these observations using methylation data of 385 patients, as part of the previously published PANTS study44-45. Nonetheless, at baseline, 323 DMPs annotated to 210 genes significantly associated with serum drug concentrations at week 14, which could potentially facilitate pre-treatment selection of individuals in need of intensified drug-monitoring. The methodological differences and lack of endoscopic data complicate direct comparison with our study. Additional analyses using endoscopically assessed anti-TNF patients is needed to provide more conclusive answers toward the utility of DNA methylation to predict response in anti-TNF treated patients.

[0078] Through two separate stability analyses, we demonstrate both short- and long-term hyper stability of the majority of our identified CpG markers indicating their independence of external exposures such as inflammatory status, which is further evidenced by the lack of correlation between the methylation of the predictor CpGs and both baseline CRP and fecal calprotectin, therapy switch and even CD-related surgery46. Further focusing on well-known confounding factors of DNA methylation signals including age, smoking behavior and cellular distribution, especially when using a mixed cell sample such as peripheral blood, we observed overall independence of the identified predictor CpGs. Notably, many of the identified predictor CpG loci annotated to genes corroborated with the mode- of-action of both drugs in literature. VDZ predictor CpGs were found to be enriched for genes involved in MHC Class 1, a pathway that has been associated primarily with self-recognition through T-cells, matching the targeted population of VDZ47-48. To that end, the MHC class 1 genes HLA-B and HLA-C as well as PIWIL1 were of particular interest49. Previous data by Verstockt et al. showed a significant upregulation of PIWIL1 in mucosal tissue associated with endoscopic remission to vedolizumab8. Notably, in our data, we observed hyper methylation in the promotor region of PIWIL1, suggesting decreased gene expression of PIWIL1 in peripheral blood as opposed to the increased mucosal expression observed by Verstockt et al. Several studies have shown that PIWIL1 stimulates, among other things, cell migration50. In hepatocellular carcinoma cells specifically, PIWI Ll-overexpression has shown to attract myeloid-derived suppressor cells in which IL-10 is expressed, mediated by PIWILl-'nduced complement C357. This potentially explains the observed difference with Verstockt et al. as decreased peripheral blood and increased mucosal expression suggests a shift of IL-10 producing suppressor cells towards the intestinal mucosa, corroborating the observed endoscopic remission by Verstockt et als. In addition to MHC Class 1, several differentially methylated genes were found to be involved in cell migration through integrin dependent celladhesion. TSPEAR, which encodes for a thrombospondin-type laminin G domain containing protein involved in integrin dependent cell-adhesion51. While 92 out of the 93 (99%) identified predictor CpGs uniquely associate with each drug, we did observe overlap of cg08017465 annotated to TSPEAR in responders to both VDZ and USTE. In addition, NID2, a type of nidogen, encodes a cell-adhesion glycoprotein, widely spread along the basement membrane. Nidogens have been reported to involve integrin dependent cell-adhesion and neutrophil chemotaxis during inflammation52.

[0079] For USTE, the difference in methylation appears more diffuse. While GO-term overrepresentation analyses did not specifically identify pathways of interest, we observed that several genes harbored predictor CpGs that are involved in macrophage- as well as Thl7 / Treg. Six DMPs could be associated with macrophage function with 4 being annotated to RHOJ, MARKS and PCGF3, which are involved in RhoA mediated IL-23 release of macrophages, while the remaining 2 DMPs are annotated to genes involved in macrophage polarization (PKN0X1 and MRC1)53 63. Furthermore, we observed differential methylation in HDAC4, a histone deacetylase previously reported to affect transcription factor AP-1 and levels of IL-6 and IL-ip involved in naive T-cell differentiation toward Thl7 cells64-68. Also, we observed overrepresentation of median-chain fatty acid transport and differential methylation in SLC27A1, which serves as a Treg fatty acid transporter known to be upregulated under hypoxic conditions to enhance Treg survival and function69-70. Further genes of interest were SMAD1 and EBF3, overrepresented as homomeric SMAD protein complex, involved in TGF-P signaling71-78. Last, we observed differential methylation in ENPP7 and enrichment of two GO-terms involved in the regulation of sphingomyelin catabolic process. ENPP7 is known as alkaline shingomyelinase, a gene involved in the hydrolyzation of shingomyeline to ceramide in the jejunum79. In rats, rectal ENPP7 administration improved DSS colitis and previous literature in psoriasis reports decreased ceramides to increase SIP, which promoted Thl7 differentiation, migration and IL-17 production80-81. These observations suggest a potential overlapping biology between ustekinumab and SIP blockage.

[0080] Our study has several major strengths to mention. First and foremost, we used a prospectively collected multi-drug cohort of CD patients, strictly phenotyped according to objectified endoscopic response combined with clinical- and biochemical parameters, providing a high level of confidence in our classification of patients into responders or non-responders. In addition, patients with anti-drug antibodies or without a measurable serum drug concentration and those that stopped treatment due to adverse events were excluded prior to the selection of this cohort. Non-responders therefore reflect a more homogenous group of true biological non-response rather than pharmacokinetic failure or intolerance. Second, the relatively small sample size was justified by performing a sample size estimation based on prior pilot experiments. Lastly, besides stability of the observed methylation differences during induction- and maintenance treatment, our markers demonstrated stable differences between R and NR over time, which was further corroborated when interrogating our 7- year longitudinal DNA methylation survey. This time-invariant behavior of the predictor CpGs indicates that exposure over time does not appear to affect the response-associated behavior of the CpGs, suggesting that the CpGs are very stable, which in turn increases its utility in clinical practice.

[0081] There are however some limitations to report as well. First of all, to ensure the development of a minimally invasive epigenetic biomarker test we used DNA derived from peripheral blood samples rather than mucosal tissue. While easily accessible and known similarities of immune cell populations for both tissues, the methylation profile of macrophages and intestinal epithelial cells are not included which may hold great significance for understanding the biological responses at the level of the mucosa. Secondly, the majority of the predictor CpG loci identified are situated within gene introns, complicating the biological interpretation of our findings. Although the identified predictor CpGs collectively serve as a strong predictor of response, we do acknowledge that the underlying biology behind this observation is more complicated than merely the inverse correlation between DNA methylation and gene expression. Lastly, this study focused on biological treatments in CD patients only. Future studies in patients with rheumatoid arthritis, atopic dermatitis, plaque psoriasis and psoriatic arthritis could further address disease or drug-specificity of the identified markers. Several US and European studies report a significant reduction in pharmacoeconomical burden of adequate treatment of IBD patients in remission compared to the cost of treating those with active IBD and suboptimal treatment82-86. In the absence of a predictive biomarker panel, current endoscopic response rates at week 52 have been reported around 45% in VDZ and 42% in USTE treated CD patients5-6. Taken together, our biomarker panels could potentially increase these proportions with 22% for VDZ and 35% for USTE, thereby significantly impacting healthcare costs and disease burden in these patients. However, clinical validation of our findings in a randomized prospective trial, comparing our method of pre-treatment selection versus current clinical practice is needed to demonstrate both clinical and cost benefit, which are both ongoing as part of the European METHYLOMIC project87.

[0082] Methods

[0083] Study population and design

[0084] We prospectively recruited adult CD patients that presented with active endoscopic disease activity upon baseline ileocolonoscopy and were scheduled to start VDZ or USTE treatment at the Amsterdam University Medical Centers, University of Amsterdam, Amsterdam, Netherlands and the John Radcliffe Hospital, Oxford, United Kingdom, which we termed the discovery and validation cohort, respectively. All patients were naive to the biological of interest.

[0085] Next, patients were treated according to standard-of-care protocols, which for VDZ meant that patients were given 300mg infusions at week 0, 2 and 6 followed by infusions at an 8 week interval and USTE treated patients received either 260-, 390- or 520mg at week 0 with subsequent 90mg s.c. injections at an 8 week interval. For both VDZ and USTE, interval intensification to either q6w or q4w as well as an extra week 10 infusion for VDZ or extra intravenous boost infusion for USTE were allowed, if needed, at the treating physicians discretion.

[0086] To ensure assessment of mechanistic and not pharmacokinetic failures to each biological, only patients with measurable serum concentrations without anti-drug antibodies at response assessment were included for analysis in this study.

[0087] The discovery cohorts were approved by the medical ethics committee of the Academic Medical Hospital (METC NL57944.018.16 and NL53989.018.15) and written informed consent was obtained from all subjects prior to sampling. The validation cohorts were approved by the National Health Service Research Ethics committee. (REC reference: 21 / PR / OO1O Protocol number: 14833 IRAS project ID: 266041) The patients were recruited and consented under the ethics of the Translational Gastrointestinal Unit biobank IBD cohort ethics (09 / H1204 / 30) and Gl cohort ethics (16 / YH / 0247 and 21 / YH / 0206). The quality of the collected data and study procedures were monitored by an independent monitor.

[0088] Sample collection and storage protocols

[0089] In both the Amsterdam discovery cohort and Oxford validation cohorts, whole peripheral blood leukocyte (PBL) samples for measurement of epigenome-wide DNA methylation were collected prior to the start of VDZ and USTE at either a baseline endoscopy- or week 0 visit (time point 1 [Tl] ) and after a median of l (20-33) weeks into treatment (time point 2 [T2] ) using 4.0-6.0mL BD ethylenediaminetetraacetic acid (EDTA) vacutainer tubes. For the Amsterdam discovery cohort, samples were subsequently aliquoted into l.lOmL micronic tubes before storing at -80gC until further handling. The Oxford validation samples were directly frozen at -80gC in preparation for later extraction.

[0090] Definitions of response

[0091] At response assessment [T2], patients were distinguished as responders (R) or non-responders (NR) based on a strict combination of endoscopic- (>50% reduction in SES-CD score) combined with either corticosteroid-free clinical- (>3 point drop88in HBI or HBI <4 and no systemic steroids) and / or biochemical response (C-reactive protein (CRP) and fecal calprotectin reduction >50% or <5 g / mL and fecal calprotectin <250 pg / g).

[0092] In the Amsterdam discovery cohorts, 6 out of 126 patients (4.8%) did not undergo endoscopic evaluation but were categorized as non-responders due to the need of surgery (n=2) or were evaluated for response using baseline- and follow-up intestinal ultrasound (n=4) with either unchanged (NR) or complete normalization of bowel wall thickness, as discussed with an expert IBD ultrasonographist (F. de Voogd).

[0093] In the Oxford validation cohorts, 42 out of 58 patients (72.4%) were evaluated using a modified definition of response where modified responders presented a combined corticosteroid-free clinical- (HBI <4) and biochemical (CRP <5 g / mL and fecal calprotectin <250 pg / g) remission between week 26-52 and remained on the same treatment at week 52. DNA isolation and in vitro DNA methylation analysis

[0094] In the Amsterdam discovery cohorts, genomic DNA was extracted using the QIAsymphony at the Amsterdam UMC, Core Facility Genomics department. We next assessed the quantity of DNA using the FLUOstar OMEGA and quality of the high-molecular weight DNA on a 0.8% agarose gel. Following these steps, 750ng of DNA per sample was randomized per plate, to limit batch effects, after which genomic DNA was bisulfite converted using the Zymo EZ DNA Methylation kit and analyzed on the Illumina HumanMethylation EPIC Beadchip array at the Core Facility Genomics, Amsterdam UMC, Amsterdam, the Netherlands.

[0095] For the oxford validation cohorts, genomic DNA was extracted using the Qiagen Puregene Blood Core Kit C at the Oxford Translational Gastroenterology Unit (Oxford). DNA samples were assessed for quality using the Nanodrop spectrometer (Nanodrop 1000, Thermo Scientific). A total of 750ng of DNA per sample was randomized per plate, to limit batch effects, after which genomic DNA was bisulfite converted using the Zymo EZ DNA Methylation kit and analyzed on the Illumina HumanMethylation EPIC Beadchip array at UCL Genomics, London, United Kingdom.

[0096] Raw methylation data pre-processing

[0097] Raw methylation data was imported into the R statistical environment using the Bioconductor minfi89'90package (version 1.36), whereupon the raw signals for the Amsterdam discovery samples were normalized using a combination of functional91- and COMBAT92normalization to minimize batch effects. We then converted the signals to methylation ratios.

[0098] Next, CpGs were removed of known genetic-variants as well as unannotated genetic variants were removed using Gaphunter93(set to 0.1) in order to increase the identification of true methylation signals rather than genetic variants.

[0099] For the validation of our models, raw methylation data from the Oxford cohorts were pre-processed and normalized together with the Amsterdam methylation data.

[0100] Machine learning models: stability selected gradient boosting analyses

[0101] To identify epigenetic markers capable of classifying responders from non-responders before the start of treatment, we used a rigorous supervised machine learning setup (figure 1A). We first split the Amsterdam discovery data into a 70% training- and 30% testing set. Notably, for each biomarker discovery iteration, the model was trained on the 70% training set without exposure to the 30% withheld testing set. Next, we used stability selection with extreme gradient boosting. Gradient boosting is a machine learning technique used for binary classification using a step-wise improvement of tree-based prediction models. Through this step-wise approach, the model builds on existing weaker trees, minimizing the overall prediction error against the observed data. The extreme gradient boosting classification algorithm optimizes a cost function by iteratively choosing a weak hypothesis that points in the negative gradient direction. Ten-fold cross-validation was used to mitigate overfitting which is a common identified problem in machine learning modelling (step la)94.

[0102] In addition, to ensure increased confidence in identifying reliable biomarker signals, we used a rigorous stability selection procedure by repeating the above mentioned steps a 100 times, each time completely reshuffling the data and applying a different split in training- and test set (step lb), as described elsewhere31. The resulting many trees (n=100), each containing its own set of ranked CpG markers according to relative importance were then combined using CID based permutation analysis

[0031]

[0103] CID based permutation importance allows us to identify the most significant biomarkers within a dataset. This method involves randomly shuffling each biomarker's values and observing the impact on the model's performance. It also allows us to better take into account various correlations within the data

[0031] , The rationale behind this is that more important biomarkers will cause a greater decrease in model performance when their values are altered. By assigning a numerical value to each biomarker and computing cumulative permutation importance based on the extent of the impact on the model's performance, we can effectively quantify their importance. This approach not only highlights key biomarkers but also allows for the exploration of various combinations (based on cumulative importance) of these markers. Consequently, it provides a deeper understanding of how different biomarkers interact and contribute to the overall quality and predictive power of the model. By assessing these combinations, we can optimize the model for better accuracy and reliability, thereby enhancing its predictive capabilities.

[0104] This test reruns the model 1000 times, each time randomly removing a CpG and permuting a random variable, assessing with each simulation the effect of this permutation on the model performance (i.e. predicted vs. true outcome). By doing so, this test ranks each CpG of the many identified trees according to its position in a joint panel of selected markers. Therefore only those CpG loci that are ranked above the performance of the permutated random variable, the predictor CpGs, are retained for future analyses (step lc). Finally, as the majority of the CpG loci identified in the many trees mentioned above are not retained due to their rank being below that of the random imputed variable, we optimized and the models by recalibrating using only the predictor CpG loci on a 80% of the discovery cohort samples and internally validated their performing using the withheld 20% discovery test set (step 2a and 2b). The resulting 50 calibrated models are then used to distinguish responders from non-responders in the Oxford external validation cohorts after which performance metrics were calculated for each model (step 3a). The average combined performance of these models in both the internal- as well as external validation (step 3b) resulted in the ROC curves shown in figure IB.

[0105] We used Python version 3.10 (www.python.org), with packages Scipy, Scikits-learn and Numpy for the model development and R version 4.2 (R Foundation, Vienna, Austria) for visualizations.

[0106] In silico DNA methylation analysis

[0107] Differential methylation analyses were performed using Umma95(version 3.46) and eBayes96regressing against age, sex, smoking behavior and estimated blood cell distribution according to Houseman's method97. Statistical significance was defined as a false discovery rate-adjusted p-value < 0.05. Visualizations were generated using ggplot298(version 3.3.5).

[0108] GO enrichment analyses

[0109] Gene ontology (GO)99-100overrepresentation analyses of the genes annotated to the response associated CpGs of interest was performed using the missMethyl package101. Gene sets with P value <0.05 were considered significant.

[0110] Sample size estimation

[0111] Based on previous pilot observations in a small set of VDZ (7 R and 5 NR) treated patients and the case-control EWAS design of this study, we estimated to need between 15-30 responder and 15-30 non-responder pairs for each biological to reach 80% EWAS power to detect significant differences, according to the method reported by Tsai and Bell102. Statistical analysis of clinical variables

[0112] Baseline characteristics of all included patients were summarized using descriptive statistics. Categorical variables are presented as percentages and continuous variables as median annotated with the interquartile range (IQR). Differences in distribution between responders, non-responders and the different cohorts were assessed using a chi-square test (categorical variables) or Mann- Whitney U (continuous variables). Two-tailed probabilities were used with a p-value of <0.05 considered as statistically significant. Analyses of clinical data were performed in IBM SPSS statistics version 26.

[0113] Tables

[0114] Table 1: Baseline characteristics discovery- and validation cohorts. Values in bold are significant, percentages shown are valid percentages. ADA: adalimumab. VDZ: vedolizumab. USTE: ustekinumab. R: responder. NR: non-responder, SD: standard deviation. IQR: interquartile range, HBI: Harvey Bradshaw Index. SES-CD: simple endoscopic disease activity score, Immunomodulator: azathioprine, mercaptopurine, thioguanine, methotrexate. Anti-TNF: infliximab, adalimumab or golimumab.

[0115] Table 2. Predictive performance metrics of the prognostic biomarkers on the train cohort acquired at the AmsterdamllMC and the validation cohort acquired at the John Radcliffe Hospital. AUROC = Area under the receiver operator characteristic. AUMC: AmsterdamllMC. JRH: John Radcliffe Hospital. VDZ: vedolizumab. USTE: ustekinumab.

[0116] Difference in Long-term

[0117] Associated methylation R vs stability Cummulative

[0118] CGID gene Chromosome Position NR (ICC) importance cgl7830959 HLA-B 6 31326324, promotor region Hypermethylation NA 15.547082 cg21070860 3 154719649 Hypermethylation 0,41 27.75169 cg08081727 12 39667364 Hypomethylation 0,98 35.34593

[0119] CTC- cg02601475 441N14.4 5 121518272, exon gene body Hypermethylation 0,73 42.876 cg00706914 TULP4 6 158735121, 1st exon Hypermethylation NA 50.11115 HLA-C / HLA- cgl7096289 B 6 31238788, introns gene body Hypomethylation 0,94 54.644955

[0120] cg25299227 ESRRB 14 76940165, intron gene body Hypermethylation 0,95 59.037754 cgO5197062 GALNT18 11 11642011, intron gene body Hypermethylation 0,98 63.075687 cg00441209 OR51A4 11 4968526, promotor region Hypomethylation NA 67.04108 cg03455316 DQ577084 15 62516405, promotor region Hypermethylation 0,96 70.86422

[0121] 31241000, intron 2 and 4 gene cg05338672 HLA-B 6 body Hypermethylation 0,97 74.62648 cgl2667521 AC005307.5 19 29218732, promotor region Hypomethylation 0,97 78.353455 cg05062694 CGREF1 2 27342324, <5kb from CGREF1 Hypomethylation NA 81.50937 cgl7764313 MCM2 3 127335263, intron gene body Hypomethylation 0,98 84.5508 cg09659072 10 114939288 Hypermethylation 0,47 87.47127 cg04674762 NID2 14 52487120, intron gene body Hypermethylation 0,95 89.93568 cg02229781 1 219104337 Hypermethylation 0,97 91.54277 cg25267487 AC005307.5 19 29217858, intron gene body Hypomethylation 0,99 92.90058 cgl6467921 8 128801108 Hypomethylation 0.97+ 94.21643 cg08017465 TSPEAR 21 46097452, intron gene body Hypermethylation 0,85 95.46908 CARNMTl- cgl4115807 AS1 9 77575609, intron gene body Hypermethylation 0,55 96.71398 cgl8319102 PIWIL1 12 130822256, promotor region Hypermethylation 0,96 97.94118 cgl2906381 RFPL2 22 32599516, intron gene body Hypermethylation 0,95 99.15726 cg04546413 AC005307.5 19 29218101, intron gene body Hypomethylation 0,99 99.759766 cg03161606 AC005307.5 19 29218774, promotor region Hypomethylation 0,98 100

[0122] Table 3. Predictor CpG loci for VDZ

[0123] Long-term

[0124] Difference in methylation R vs stability

[0125] CGID Position NR (ICC) Cumulativejmportanc

[0126] 20376453 (1.5Kb from 1st exon cgl3982436 22 TMEM191B) Hypomethylation NA 10.507598 cgll079896 14 63671314, 1st exon Hypermethylation 0,98 20.737144 cg09147516 6 43365977 (>20kb from 1st exon ZNF318) Hypomethylation 0,91 29.733046 cgl9162470 15 31192891 (3kb from 1st exon FAN1) Hypomethylation NA 36.267513 cg05541470 12 132061506 (>100kb from 1st exon SFSWAP) Hypermethylation 0,92 41.45134 cgl3754978 8 1412532, intron gene body Hypermethylation NA 46.61049 cgl8771300 14 63671737, 1st exon Hypermethylation 0,98 51.49284 cgl3816228 1 90022780, gene body (intron) Hypomethylation 0,74 56.11453 cg08270491 8 126100743, intron gene body Hypomethylation 0,95 60.11879 cgl4578009 5 180597642 (lkb from 1st exon tRNA_Val ) Hypermethylation 0,95 63.382 cg01966334 2 128378434, intron gene body Hypomethylation 0,93 66.45093 cg02265440 14 70690079, intron gene body Hypermethylation NA 69.48873 cgl7190362 6 112475675, gene body (intron) Hypermethylation 0,98 72.145 cgl3468767 17 9672024 (3 kb from 6th exon DHRS7C) Hypermethylation 0,94 74.72114 cgl7132030 19 17599784, intron gene body Hypomethylation 0,89 77.09166 cg20434511 2 240240975, gene body (intron) Hypomethylation 0,94 79.28673 cg23973310 2 65955240, gene body (intron) Hypomethylation 0,94 81.287605 cg25279747 6 43354368 (15kb from 1st exon ZNF318) Hypermethylation 0,97 83.23028 cgll747594 6 29648225, gene body (intron) Hypermethylation 0,98 85.042946 cgl7481116 8 14122001, intron gene body Hypermethylation 0,95 86.83479 cg02086964 19 54941218, gene body (intron) Hypermethylation 0,97 88.466484 cglll41652 22 24348549, promotor region (1291bp) Hypomethylation 0,77 89.68711 cgl2119625 19 54106789, promotor region Hypomethylation 0,99 90.72723 cgl5921713 7 16602973, gene body (intron) Hypermethylation 0,93 91.76347 cg21650737 8 674525, intron gene body Hypermethylation 0,84 92.628685 cgl4829155 15 31115871, exon 3 gene body Hypermethylation 0,98 93.38718 cg07620573 3 192289293, intron gene body Hypermethylation 0,94 94.12589

[0127] cg25535666 2 201480219, gene body (intron) Hypomethylation 0,76 94.85411 cg02860608 3 134084216 (>10kb from 1st exon AMOTL2) Hypermethylation 0,87 95.54217 105343122 (>17kb from 1st exon cg20707527 8 DCSTAMP) Hypomethylation 0,98 96.12128 cgl4157578 9 110759967 (>400kb from 1st exon KLF4) Hypermethylation 0,95 96.538475 cgl7037048 12 106100507, gene body (intron) Hypermethylation NA 96.9419 cg27408471 15 73940201 (>14kb from 1st exon NPTN) Hypermethylation 0,95 97.26248 cg04996388 17 79905263, promotor region MYADML2 Hypomethylation NA 97.54137 cg22635676 2 241975971, gene body (intron) Hypermethylation 0,93 97.79523 cg24309011 21 44444757, gene body (intron) Hypermethylation NA 98.04427 cg08359343 4 146478783, intron gene body Hypermethylation NA 98.27548 cg04043455 10 131669461, intron gene body Hypermethylation 0,94 98.50622 cgl6677969 10 85677559 (200kb from 1st exon GHITM) Hypomethylation 0,99 98.71156 cg06546677 8 145537504, gene body (intron) Hypomethylation 0,98 98.89881 cgl7851604 2 202753217, gene body (exon) Hypermethylation 0,94 99.05759 cgl4024893 9 139943146, gene body (exon) Hypermethylation NA 99.20722 cgl9072128 2 239337689, gene body (intron) Hypermethylation 0,64 99.32528 cg22329875 8 55421152 (>40kb from SOX17) Hypermethylation 0,98 99.43581 cgll935738 6 29520752, gene body (intron) Hypomethylation 0,92 99.540764

[0128] 139539001, promotor region NXPH2 cg08010094 2 (1190bp) Hypomethylation 0,98 99.635544 cg23264413 19 43710277, promotor region PSG4 Hypomethylation 0,95 99.68395 cglll46691 12 47219737, 1st exon SLC38A4 Hypermethylation 0,95 99.723305 cgl3746813 6 14911904 (>500kb from 1st exon JARID2) Hypomethylation 0,88 99.76234 cg05303293 13 25052281, gene body (exon) Hypomethylation 0,9 99.80024 cg24536782 8 216659 (>20kb from ZNF596) Hypermethylation NA 99.82356 cgl0167378 1 228756711 (500bp from 1st exon RNA5S6) Hypomethylation 0,89 99.84555 cg23762517 1 42384310, 1st exon HIVEP3 Hypomethylation 0,91 99.86491

[0129] cg08109568 15 31115862, gene body (exon) Hypermethylation 0,97 99.883095 cgll787544 13 29257933 (5kb from 6th exon POMP) Hypermethylation 0,5 99.900444 cg00876837 10 18155409, gene body (intron) Hypomethylation 0,65 99.91327 cg06880335 14 103894049, gene body (intron) Hypomethylation 0,86 99.92607 cgl9678447 1 31922081 (>14kb from last exon SERINC2) Hypermethylation 0,87 99.93864 cg08017465 21 2146097452, gene body (intron) Hypomethylation 0,85 99.949394 cg08993878 12 98151379 (>lkb from 1st exon LOC643711) Hypomethylation 0,97 99.95928 cg02094681 5 113661575, gene body (intron) Hypomethylation 0,97 99.96707 cg20171775 2 145228686, gene body (intron) Hypermethylation 0,93 99.974335 cgl0403394 15 63349192, gene body (exon) Hypomethylation 0,82 99.9811 cg08777654 15 22548405, gene body (intron) Hypermethylation NA 99.98717 cgl3679714 17 77706946, gene body (intron) Hypermethylation 0,98 99.990906 cg09125754 2 130886714, 1st exon POTEF Hypomethylation 0,95 99.99415 cg23730027 3 57995180, gene body (intron) Hypomethylation NA 99.99737 cgl0864200 4 720809, gene body (intron) Hypomethylation 0,91 99.99999

[0130] Table 4. Predictor CpG loci for USTE

[0131] Baseline characteristics multi-drug failures N=34

[0132] Gender, n (%), Female 18 (52.9)

[0133] Age, years, median (IQR) 36 (29-46)

[0134] Disease duration, years, median (IQR) 13 (9-20)

[0135] Ethnic background, n (%), Caucasian 24 (70.6)

[0136] Disease location, n (%)

[0137] Ileal disease (LI) 3 (8.8)

[0138] Colonic disease (L2) 8 (23.5)

[0139] Ileocolonic disease (L3) 23 (67.6)

[0140] Disease behavior, n (%)

[0141] Non structuring non-penetrating (Bl) 10 (29.4)

[0142] Stricturing (B2) 11 (32.4)

[0143] Penetrating (B3) 13 (38.2)

[0144] Perianal disease (p) 18 (52.9)

[0145] Previous IBD related surgery, n (%) 24 (70.6)

[0146] Previous objectified treatment failure, n (%) anti-TNF, vedolizumab and ustekinumab 14 (41.2) anti-TNF and vedolizumab 8 (23.5) anti-TNF and ustekinumab 7 (20.6) vedolizumab and ustekinumab 5 (14.7)

[0147] Smoking, n (%)d

[0148] Non-smoker 31 (91.2)

[0149] Active smoker 3 (8.8)

[0150] Table 5: Clinical characteristics multi-biologic failure patients

[0151] References

[0152] 1. permutationde Souza HS, Fiocchi C. Immunopathogenesis of IBD: current state of the art. Nat Rev Gastroenterol Hepatol 2016;13:13-27.

[0153] 2. Liu JZ, van Sommeren S, Huang H, et al. Association analyses identify 38 susceptibility loci for inflammatory bowel disease and highlight shared genetic risk across populations. Nat Genet 2015;47:979-986.

[0154] 3. Alsoud D, Vermeire S, Verstockt B. Biomarker discovery for personalized therapy selection in inflammatory bowel diseases: Challenges and promises. Curr Res Pharmacol Drug Discov 2022;3:100089.

[0155] 4. Juillerat P, Grueber MM, Ruetsch R, et al. Positioning biologies in the treatment of IBD: A practical guide - Which mechanism of action for whom? Curr Res Pharmacol Drug Discov 2022;3:100104.

[0156] 5. Lowenberg M, Vermeire S, Mostafavi N, et al. Vedolizumab Induces Endoscopic and Histologic Remission in Patients With Crohn's Disease. Gastroenterology 2019;157:997- 1006 e6.

[0157] 6. Sands BE, Irving PM, Hoops T, et al. Ustekinumab versus adalimumab for induction and maintenance therapy in biologic-naive patients with moderately to severely active Crohn's disease: a multicentre, randomised, double-blind, parallel-group, phase 3b trial. Lancet 2022;399:2200-2211.

[0158] 7. Colombel JF, Sandborn WJ, Reinisch W, et al. Infliximab, azathioprine, or combination therapy for Crohn's disease. N Engl J Med 2010;362:1383-95.

[0159] 8. Verstockt B, Verstockt S, Veny M, et al. Expression Levels of 4 Genes in Colon Tissue Might Be Used to Predict Which Patients Will Enter Endoscopic Remission After Vedolizumab Therapy for Inflammatory Bowel Diseases. Clin Gastroenterol Hepatol 2020;18:1142-1151 elO.

[0160] 9. Verstockt B, Verstockt S, Dehairs J, et al. Low TREMl expression in whole blood predicts anti-TNF response in inflammatory bowel disease. EBioMedicine 2019;40:733-742.

[0161] 10. Verstockt B, Sudahakar P, Creyns B, et al. DOP70 An integrated multi-omics biomarker predicting endoscopic response in ustekinumab treated patients with Crohn's disease. Journal of Crohn's and Colitis 2019;13:S072-S073.

[0162] 11. Soendergaard C, Seidelin JB, Steenholdt C, et al. Putative biomarkers of vedolizumab resistance and underlying inflammatory pathways involved in IBD. BMJ Open Gastroenterol 2018;5:e000208.

[0163] 12. Ananthakrishnan AN, Luo C, Yajnik V, et al. Gut Microbiome Function Predicts Response to Anti-integrin Biologic Therapy in Inflammatory Bowel Diseases. Cell Host Microbe 2017;21:603-610 e3.

[0164] 13. Lee JWJ, Plichta D, Hogstrom L, et al. Multi-omics reveal microbial determinants impacting responses to biologic therapies in inflammatory bowel disease. Cell Host Microbe 2021;29:1294-1304 e4.

[0165] 14. Stevens TW, Matheeuwsen M, Lonnkvist MH, et al. Systematic review: predictive biomarkers of therapeutic response in inflammatory bowel disease-personalised medicine in its infancy. Aliment Pharmacol Ther 2018;48:1213-1231.

[0166] 15. Gisbert JP, Chaparro M. Predictors of Primary Response to Biologic Treatment [Anti-TNF, Vedolizumab, and Ustekinumab] in Patients With Inflammatory Bowel Disease: From Basic Science to Clinical Practice. J Crohns Colitis 2020;14:694-709.

[0167] 16. Zilbauer M, Heuschkel R. Disease Prognostic Biomarkers in Inflammatory Bowel Diseases-A Reality Check. J Crohns Colitis 2022;16:162-165.

[0168] 17. Bird A. DNA methylation patterns and epigenetic memory. Genes Dev 2002;16:6-21. 18. Kaluscha S, Domcke S, Wirbelauer C, et al. Evidence that direct inhibition of transcription factor binding is the prevailing mode of gene and repeat repression by DNA methylation. Nat Genet 2022;54:1895-1906.

[0169] 19. de Mendoza A, Nguyen TV, Ford E, et al. Large-scale manipulation of promoter DNA methylation reveals context-specific transcriptional responses and stability. Genome Biol 2022;23:163.

[0170] 20. Hornschuh M, Wirthgen E, Wolfien M, et al. The role of epigenetic modifications for the pathogenesis of Crohn's disease. Clin Epigenetics 2021;13:108.

[0171] 21. Peery RC, Pammi M, Claud E, et al. Epigenome - A mediator for host-microbiome crosstalk. Semin Perinatol 2021;45:151455.

[0172] 22. Qin Y, Wade PA. Crosstalk between the microbiome and epigenome: messages from bugs. J Biochem 2018;163:105-112.

[0173] 23. Li Y, Wang Z, Wu X, et al. Intestinal mucosa-derived DNA methylation signatures in the penetrating intestinal mucosal lesions of Crohn's disease. Sci Rep 2021;ll:9771.

[0174] 24. Sadler T, Bhasin JM, Xu Y, et al. Genome-wide analysis of DNA methylation and gene expression defines molecular characteristics of Crohn's disease-associated fibrosis. Clin Epigenetics 2016;8:30.

[0175] 25. Adams AT, Kennedy NA, Hansen R, et al. Two-stage Genome-wide Methylation Profiling in Childhood-onset Crohn's Disease Implicates Epigenetic Alterations at the VMP1 / MIR21 and HLA Loci. Inflammatory Bowel Diseases 2014;20:1784-1793.

[0176] 26. Ventham NT, Kennedy NA, Adams AT, et al. Integrative epigenome-wide analysis demonstrates that DNA methylation may mediate genetic risk in inflammatory bowel disease. Nature Communications 2016;7:13507.

[0177] 27. Li Yim AYF, Duijvis NW, Zhao J, et al. Peripheral blood methylation profiling of female Crohn's disease patients. Clin Epigenetics 2016;8:65.

[0178] 28. Howell KJ, Kraiczy J, Nayak KM, et al. DNA Methylation and Transcription Patterns in Intestinal Epithelial Cells From Pediatric Patients With Inflammatory Bowel Diseases Differentiate Disease Subtypes and Associate With Outcome. Gastroenterology 2018;154:585-598.

[0179] 29. Gasparetto M, Payne F, Nayak K, et al. Transcription and DNA Methylation Patterns of Blood-Derived CD8(+) T Cells Are Associated With Age and Inflammatory Bowel Disease But Do Not Predict Prognosis. Gastroenterology 2020.

[0180] 30. Li Yim AYF, Duijvis NW, Ghiboub M, et al. Whole-Genome DNA Methylation Profiling of CD14+ Monocytes Reveals Disease Status and Activity Differences in Crohn's Disease Patients. J Clin Med 2020;9.

[0181] 31. J. P, E. S, A. Z, et al. Covered Information Disentanglement: Model Transparency via Unbiased Permutation Importance. The Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI-22), 2022.

[0182] 32. Joustra V. LYA. Long-term temporal stability of peripheral blood DNA methylation alterations in patients with inflammatory bowel disease. BioRxiv, 2022.

[0183] 33. Tsai PC, Glastonbury CA, Eliot MN, et al. Smoking induces coordinated DNA methylation and gene expression changes in adipose tissue with consequences for metabolic health. Clin Epigenetics 2018;10:126.

[0184] 34. Houseman EA, Kelsey KT, Wiencke JK, et al. Cell-composition effects in the analysis of DNA methylation array data: a mathematical perspective. BMC Bioinformatics 2015;16:95.

[0185] 35. Solomon O, Huen K, Yousefi P, et al. Meta-analysis of epigenome-wide association studies in newborns and children show widespread sex differences in blood DNA methylation. Mutation Research-Reviews in Mutation Research 2022;789.

[0186] 36. Dobbs KR, Embury P, Koech E, et al. Age-related differences in monocyte DNA methylation and immune function in healthy Kenyan adults and children. Immun Ageing 2021;18:ll. 37. Somineni HK, Venkateswaran S, Kilaru V, et al. Blood-Derived DNA Methylation Signatures of Crohn Disease and Severity of Intestinal Inflammation. Gastroenterology 2019.

[0187] 38. Kappelman MD, Adimadhyam S, Hou L, et al. Real-World Evidence Comparing Vedolizumab and Ustekinumab in Antitumor Necrosis Factor-Experienced Patients With Crohn's Disease. Am J Gastroenterol 2023;118:674-684.

[0188] 39. Torres J, Bonovas S, Doherty G, et al. ECCO Guidelines on Therapeutics in Crohn's Disease: Medical Treatment. J Crohns Colitis 2020;14:4-22.

[0189] 40. Siegel C, Sebastian S, Siegmund B, et al. P714 Mucosal healing in CD with vedolizumab versus other biologies: endoscopic outcomes during long-term routine care in a multinational observational study. Journal of Crohn's and Colitis 2023;17:i844-i844.

[0190] 41. Danese S, Vermeire S, D'Haens G, et al. Treat to target versus standard of care for patients with Crohn's disease treated with ustekinumab (STARDUST): an open-label, multicentre, randomised phase 3b trial. Lancet Gastroenterol Hepatol 2022;7:294-306.

[0191] 42. Feagan BG, Lasch K, Lissoos T, et al. Rapid Response to Vedolizumab Therapy in Biologic- Naive Patients With Inflammatory Bowel Diseases. Clin Gastroenterol Hepatol 2019;17:130-138 e7.

[0192] 43. Mishra N, Aden K, Blase JI, et al. Longitudinal multi-omics analysis identifies early bloodbased predictors of anti-TNF therapy response in inflammatory bowel disease. Genome Med 2022;14:110.

[0193] 44. Lin S, Hannon E, Reppell M, et al. Whole blood DNA methylation changes are associated with anti-TNF drug concentration in patients with Crohn's disease. J Crohns Colitis 2023.

[0194] 45. Kennedy NA, Heap GA, Green HD, et al. Predictors of anti-TNF treatment failure in anti- TNF-naive patients with active luminal Crohn's disease: a prospective, multicentre, cohort study. Lancet Gastroenterol Hepatol 2019;4:341-353.

[0195] 46. Joustra V, Li Yim AYF, Hageman I, et al. Long-term Temporal Stability of Peripheral Blood DNA Methylation Profiles in Patients With Inflammatory Bowel Disease. Cell Mol Gastroenterol Hepatol 2023;15:869-885.

[0196] 47. Berlin C, Berg EL, Briskin MJ, et al. Alpha 4 beta 7 integrin mediates lymphocyte binding to the mucosal vascular addressin MAdCAM-1. Cell 1993;74:185-95.

[0197] 48. Wyant T, Fedyk E, Abhyankar B. An Overview of the Mechanism of Action of the Monoclonal Antibody Vedolizumab. J Crohns Colitis 2016;10:1437-1444.

[0198] 49. Parham P, Moffett A. Variable NK cell receptors and their MHC class I ligands in immunity, reproduction and human evolution. Nature Reviews Immunology 2013;13:133-144.

[0199] 50. Dong P, Xiong Y, Konno Y, et al. Critical Roles of PIWIL1 in Human Tumors: Expression, Functions, Mechanisms, and Potential Clinical Implications. Front Cell Dev Biol 2021;9:656993.

[0200] 51. Ido H, Harada K, Futaki S, et al. Molecular dissection of the alpha-dystroglycan- and integrin-binding sites within the globular domain of human laminin-10. Journal of Biological Chemistry 2004;279:10946-10954.

[0201] 52. Ulazzi L, Sabbioni S, Miotto E, et al. Nidogen 1 and 2 gene promoters are aberrantly methylated in human gastrointestinal cancer. Mol Cancer 2007;6:17.

[0202] 53. Yuan L, Sacharidou A, Stratman AN, et al. RhoJ is an endothelial cell-restricted Rho GTPase that mediates vascular morphogenesis and is regulated by the transcription factor ERG. Blood 2011;118:1145-53.

[0203] 54. Montaner S, Perona R, Saniger L, et al. Activation of serum response factor by RhoA is mediated by the nuclear factor-kappaB and C / EBP transcription factors. J Biol Chem 1999;274:8506-15.

[0204] 55. Segain JP, Raingeard de la Bletiere D, Sauzeau V, et al. Rho kinase blockade prevents inflammation via nuclear factor kappa B inhibition: evidence in Crohn's disease and experimental colitis. Gastroenterology 2003;124:1180-7. 56. Yuan L, Sacharidou A, Stratman AN, et al. RhoJ is an endothelial cell-restricted Rho GTPase that mediates vascular morphogenesis and is regulated by the transcription factor ERG. Blood 2011;118:1145-1153.

[0205] 57. Park SY, Lee SW, Lee WS, et al. RhoA / ROCK-dependent pathway is required for TLR2- mediated IL-23 production in human synovial macrophages: Suppression by cilostazol. Biochemical Pharmacology 2013;86:1320-1327.

[0206] 58. Zeng RJ, Zhuo ZW, Luo YJ, et al. Rho GTPase signaling in rheumatic diseases. Iscience 2022;25.

[0207] 59. Sandi MJ, Marshall CB, Balan M, et al. MARK3-mediated phosphorylation of ARHGEF2 couples microtubules to the actin cytoskeleton to establish cell polarity. Sci Signal 2017;10.

[0208] 60. Hu Y, Cheng Y, Jiang X, et al. PCGF3 promotes the proliferation and migration of non-small cell lung cancer cells via the PI3K / AKT signaling pathway. Exp Cell Res 2021;400:112496.

[0209] 61. Zhuang GQ, Meng C, Guo X, et al. A Novel Regulator of Macrophage Activation miR-223 in Obesity-Associated Adipose Tissue Inflammation. Circulation 2012;125:2892-+.

[0210] 62. Jiao P, Wang XP, Luoreng ZM, et al. miR-223: An Effective Regulator of Immune Cell Differentiation and Inflammation. International Journal of Biological Sciences 2021;17:2308-2322.

[0211] 63. Wright PB, McDonald E, Bravo-Blas A, et al. The mannose receptor (CD206) identifies a population of colonic macrophages in health and inflammatory bowel disease. Sci Rep 2021;ll:19616.

[0212] 64. Lu W, You R, Yuan XY, et al. The microRNA miR-22 inhibits the histone deacetylase HDAC4 to promote T(H)17 cell-dependent emphysema. Nature Immunology 2015;16:1185-1194.

[0213] 65. Glauben R, Sonnenberg E, Wetzel M, et al. Histone Deacetylase Inhibitors Modulate Interleukin 6-dependent CD4(+) T Cell Polarization in Vitro and in Vivo. Journal of Biological Chemistry 2014;289:6142-6151.

[0214] 66. Brusselle GG, Bracke KR. MicroRNA miR-22 drives T(H)17 responses in emphysema. Nature Immunology 2015;16:1109-1110.

[0215] 67. Dou B, Ma FZ, Jiang ZY, et al. Blood HDAC4 Variation Links With Disease Activity and Response to Tumor Necrosis Factor Inhibitor and Regulates CD4+T Cell Differentiation in Ankylosing Spondylitis. Frontiers in Medicine 2022;9.

[0216] 68. Brusselle GG, Bracke KR. MicroRNA miR-22 drives T(H)17 responses in emphysema. Nat Immunol 2015;16:1109-10.

[0217] 69. Miska J, Lee-Chang C, Rashidi A, et al. HIF-1 alpha Is a Metabolic Switch Between Glycolytic-Driven Migration and Oxidative Phosphorylation-Driven Immunosuppression of Tregs in Glioblastoma (vol 27 , 226-237. el-e4, 2019). Cell Reports 2022;39.

[0218] 70. Gao R, Shi GP, Wang J. Functional Diversities of Regulatory T Cells in the Context of Cancer Immunotherapy. Front Immunol 2022;13:833667.

[0219] 71. Browning LM, Pietrzak M, Kuczma M, et al. TGF-beta-mediated enhancement of TH17 cell generation is inhibited by bone morphogenetic protein receptor lalpha signaling. Sci Signal 2018;ll.

[0220] 72. Kuczma M, Kraj P. Bone Morphogenetic Protein Signaling Regulates Development and Activation of CD4(+) T Cells. Bone Morphogenic Protein 2015;99:171-193.

[0221] 73. Lu L, Ma JL, Wang XH, et al. Synergistic effect of TGF-beta superfamily members on the induction of Foxp3(+) Treg. European Journal of Immunology 2010;40:142-152.

[0222] 74. Park SH, Kim SK, Choe JY, et al. Hypermethylation of EBF3 and IRX1 genes in synovial fibroblasts of patients with rheumatoid arthritis. Molecules and Cells 2013;35:298-304.

[0223] 75. Stockinger B, Veldhoen M. Differentiation and function of Thl7 T cells. Current Opinion in Immunology 2007;19:281-286.

[0224] 76. Lyakh L, Trinchieri G, Provezza L, et al. Regulation of interleukin-12 / interleukin-23 production and the T-helper 17 response in humans. Immunological Reviews 2008;226:112-131. 77. Zhao L, Ghetie D, Jiang L, et al. How Can We Manipulate the I L-23 / 1 L-17 Axis? Current Treatment Options in Rheumatology 2015;1:182-196.

[0225] 78. Morishima N, Mizoguchi I, Takeda K, et al. TGF-beta is necessary for induction of IL-23R and Thl7 differentiation by IL-6 and IL-23. Biochemical and Biophysical Research Communications 2009;386:105-110.

[0226] 79. Duan RD, Bergman T, Xu N, et al. Identification of human intestinal alkaline sphingomyelinase as a novel ecto-enzyme related to the nucleotide phosphodiesterase family. J Biol Chem 2003;278:38528-36.

[0227] 80. Zhang P, Chen Y, Zhang T, et al. Deficiency of alkaline SMase enhances dextran sulfate sodium-induced colitis in mice with upregulation of autotaxin. J Lipid Res 2018;59:1841- 1850.

[0228] 81. Li Q, Fang H, Dang E, et al. The role of ceramides in skin homeostasis and inflammatory skin diseases. J Dermatol Sci 2020;97:2-8.

[0229] 82. Yu AP, Cabanilla LA, Wu EQ, et al. The costs of Crohn's disease in the United States and other Western countries: a systematic review. Curr Med Res Opin 2008;24:319-28.

[0230] 83. Mehta F. Report: economic implications of inflammatory bowel disease and its management. Am J Manag Care 2016;22:s51-60.

[0231] 84. Jackson B, Con D, Ma R, et al. Health care costs associated with Australian tertiary inflammatory bowel disease care. Scand J Gastroenterol 2017;52:851-856.

[0232] 85. Ghosh N, Premchand P. A UK cost of care model for inflammatory bowel disease. Frontline Gastroenterol 2015;6:169-174.

[0233] 86. Rubin DT, Mody R, Davis KL, et al. Real-world assessment of therapy changes, suboptimal treatment and associated costs in patients with ulcerative colitis or Crohn's disease. Aliment Pharmacol Ther 2014;39:1143-55.

[0234] 87. DNA methylation markers to predict treatment success of biologicals in Crohn's disease.

[0235] 88. Vermeire S, Schreiber S, Sandborn WJ, et al. Correlation between the Crohn's disease activity and Harvey-Bradshaw indices in assessing Crohn's disease severity. Clin Gastroenterol Hepatol 2010;8:357-63.

[0236] 89. Aryee MJ, Jaffe AE, Corrada-Bravo H, et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics 2014;30:1363-9.

[0237] 90. Fortin JP, Triche TJ, Jr., Hansen KD. Preprocessing, normalization and integration of the Illumina HumanMethylationEPIC array with minfi. Bioinformatics 2017;33:558-560.

[0238] 91. Fortin JP, Labbe A, Lemire M, et al. Functional normalization of 450k methylation array data improves replication in large cancer studies. Genome Biol 2014;15:503.

[0239] 92. Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 2007;8:118-27.

[0240] 93. Andrews SV, Ladd-Acosta C, Feinberg AP, et al. "Gap hunting" to characterize clustered probe signals in Illumina methylation array data. Epigenetics Chromatin 2016;9:56.

[0241] 94. Demsar J, Zupan B. Hands-on training about overfitting. PLoS Comput Biol 2021;17:el008671.

[0242] 95. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA- sequencing and microarray studies. Nucleic Acids Res 2015;43:e47.

[0243] 96. Smyth GK. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 2004;3:Article3.

[0244] 97. Houseman EA, Accomando WP, Koestler DC, et al. DNA methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinformatics 2012;13:86.

[0245] 98. Wickham H. ggplot2: Elegant Graphics for Data Analysis: Springer-Verlag New York, 2016.

[0246] 99. Ashburner M, Ball CA, Blake JA, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 2000;25:25-9. 100. Gene Ontology C. The Gene Ontology resource: enriching a GOId mine. Nucleic Acids Res 2021;49:D325-D334.

[0247] 101. Phipson B, Maksimovic J, Oshiack A. miss Methyl: an R package for analyzing data from Illumina's HumanMethylation450 platform. Bioinformatics 2016;32:286-8. 102. Tsai PC, Bell JT. Power and sample size estimation for epigenome-wide association scans to detect differential DNA methylation. Int J Epidemiol 2015;44:1429-1441.

[0248] 103. Koo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. J Chiropr Med 2016;15:155-63.

Claims

CLAIMS1. A method of determining or predicting the sensitivity of a subject to an antiinflammatory treatment against IBD using vedolizumab, comprising the steps of: a. Providing a biological sample of a subject suffering from IBD, b. determining the methylation status of at least one CpG selected from the group consisting of cgl7830959, cg21070860, cg08081727, cg02601475, cg00706914, cgl7096289, cg25299227, cg05197062, cg00441209, cg03455316, cg05338672, cgl2667521, cg05062694, cgl7764313, cg09659072, cg04674762, cg02229781, cg25267487, cgl6467921, cg08017465, cgl4115807, cgl8319102, cgl2906381, cg04546413, cg03161606, and c. determining the sensitivity based on said methylation status wherein a higher level of methylation of cgl7830959, cg21070860, cg02601475, cg00706914, cg25299227, cg05197062, cg03455316, cg05338672, cg09659072, cg04674762, cg02229781, cg08017465, cgl4115807, cgl8319102, cgl2906381, and a lower level of methylation of cg08081727, cgl7096289, cg00441209, cgl2667521, cg05062694, cgl7764313, cg25267487, cgl6467921, cg04546413, cg03161606 in comparison to a control value or control sample is indicative of an increased sensitivity to a therapy using vedolizumab.

2. A method of determining or predicting the sensitivity of a subject to an antiinflammatory treatment against IBD using Ustekinumab, comprising the steps of: a. Providing a biological sample of a subject suffering from IBD, b. determining the methylation status of at least one CpG selected from the group consisting of cgl3982436, cgl 1079896, cg09147516, cgl9162470, cg05541470, cgl3754978, cgl8771300, cgl3816228, cg08270491, cgl4578009, cg01966334, cg02265440, cgl7190362, cgl3468767, cgl7132030, cg20434511, cg23973310, cg25279747, cgl 1747594, cgl7481116, cg02086964, cgl 1141652, cgl2119625, cgl5921713, cg21650737, cgl4829155, cg07620573, cg25535666, cg02860608, cg20707527, cgl4157578, cgl7037048, cg27408471, cg04996388, cg22635676, cg24309011, cg08359343, cg04043455, cgl6677969, cg06546677, cgl7851604, cgl4024893, cgl9072128, cg22329875, cgl 1935738, cg08010094, cg23264413, cgl 1146691, cgl3746813, cg05303293, cg24536782, cgl0167378, cg23762517, cg08109568, cgl 1787544, cg00876837, cg06880335, cgl9678447, cg08017465, cg08993878, cg02094681, cg20171775, cgl0403394, cg08777654, cgl3679714, cg09125754, cg23730027, cgl0864200, andc. determining the sensitivity based on said methylation status wherein a higher level of methylation of cgl 1079896, cg05541470, cgl3754978, cgl8771300, cgl4578009, cg02265440, cgl7190362, cgl3468767, cg25279747, cgl 1747594, cgl7481116, cg02086964, cgl5921713, cg21650737, cgl4829155, cg07620573, cg02860608, cgl4157578, cgl7037048, cg27408471, cg22635676, cg24309011, cg08359343, cg04043455, cgl7851604, cgl4024893, cgl9072128, cg22329875, cgl 1146691, cg24536782, cg08109568, cgl 1787544, cgl9678447, cg20171775, cg08777654, cgl3679714, and a lower level of methylation of cgl3982436, cg09147516, cgl9162470, cgl3816228, cg08270491, cg01966334, cgl7132030, cg20434511, cg23973310, cgl 1141652, cgl2119625, cg25535666, cg20707527, cg04996388, cgl6677969, cg06546677, cgl 1935738, cg08010094, cg23264413, cgl3746813, cg05303293, cgl0167378, cg23762517, cg00876837, cg06880335, cg08017465, cg08993878, cg02094681, cgl0403394, cg09125754, cg23730027, eg 10864200 in comparison to a control value or control sample is indicative of an increased sensitivity to a therapy using Ustekinumab.

3. Method according to claim 1 or 2, wherein said IBD is Crohn’s disease.

4. Method according to any of claims 1 - 3, wherein said biological sample comprises white blood cells.

5. Method according to any of claims 1 - 4, wherein said methylation level is determined using DNA methylation array.

6. Method according to any of claims 1, 3-5, wherein at least one CpG comprises 2, more preferably 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or all CpGs of Table 3.

7. Method according to any of claims 1, 3-6, wherein at least one CpG comprises the first 2, more preferably the first 3, 4, 5, 6, 7.

8.

9. 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or all CpGs of Table 3.

8. Method according to any of claims 2-5, wherein at least one CpG comprises 2, more preferably 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67 or all CpG of Table 4.

9. Method according to any of claims 2-5 and 8, wherein at least one CpG comprises the first CpG first 2, more preferably the first 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, or all CpG of Table 4.

10. Vedolizumab for use in the treatment of IBD in a subject who is sensitive to a treatment with vedolizumab as determined using the method according to any of claims 1, 3-7.

11. Ustekinumab for use in the treatment of IBD in a subject who is sensitive to a treatment with Ustekinumab as determined using the method according to any of claims 2-5, 8 and 9.

12. Method according to claim 1-9, wherein said subject was a CD patient having a treatment failure as may be determined through endoscopy

Images